Amazon Translate

What Amazon Translate does

Amazon Translate is an AWS service for translating text and supported documents between languages. Developers can put translation inside an application, while teams with larger document collections can run batch jobs. It suits multilingual support content, product descriptions, internal knowledge articles and other written material that needs a repeatable translation process.

The developer guide also describes connections to other AWS services. Those are separate parts of a workflow: translating a transcript requires text first, and creating spoken output requires a speech service. Translate itself provides the text translation stage.

Real-time requests and document batches

Real-time translation accepts UTF-8 text or a supported document request. AWS documents TXT, HTML and DOCX for real-time file translation. For asynchronous work, files are placed in Amazon S3 and submitted as a batch job; supported batch formats include TXT, HTML, DOCX, XLSX, PPTX and XLIFF 1.2. The service returns translated files in the corresponding format.

These two routes serve different needs. An application can translate a customer message when it arrives. A localization team can process a collection of manuals without sending each document interactively. The request and format documentation is useful when choosing the route and preparing source files.

Controls for terminology and style

Custom terminology lets a team supply preferred translations for specific terms, such as a product name or specialist vocabulary. Parallel data supplies examples of source text and its desired translation for Active Custom Translation. These are different controls: a terminology list addresses chosen words, while paired examples help adapt how passages are translated.

Other documented options include formality, profanity masking, brevity and HTML markers for text that should remain untranslated. Support depends on the language and request type, so check the customization guide before designing an interface that promises a setting for every language.

An example localization workflow

A software company could start with a reviewed English help article and a list of approved product terms. A developer sends the article to Translate with the selected source and target languages, then stores the response as a draft. A reviewer who understands the target language checks instructions, links and terminology before the article is published.

For a quarterly documentation release, the team could instead gather supported documents in S3 and submit a batch job. Keep source versions and translated versions associated so reviewers can trace which original a translation came from.

Limitations and AWS access

Machine translation still needs review for ambiguous wording, specialized instructions and consequential material. Automatic source-language detection uses Amazon Comprehend; AWS documents detection-confidence and unsupported-language errors. Specify a known source language when practical, and confirm the required language pair and file format before building a production workflow.

Amazon Translate requires AWS access and charges according to usage. Its pricing page distinguishes translation types and counts translated characters, including whitespace. Free-tier eligibility and allowances have conditions; they do not make all translation features free. Budget separately for associated services such as S3 storage or language detection, and check the current rates for the request type you intend to use.